Secreted C-type lectin regulation of neuromuscular junction synaptic vesicle dynamics modulates coordinated movement

Synapse assembly and subsequent neurotransmission strength are dependent on secreted bidirectional signaling pathways, which communicate between presynaptic and postsynaptic partners to coordinate synaptic development, function and plasticity (Rushton et al., 2020). Both synaptic cleft and the surrounding extracellular perisynaptic space (the ‘synaptomatrix’) are characterized by heavy glycosylation, which has been shown over the last half-century with extensive lectin imaging (Dani and Broadie, 2012; Dani et al., 2012; Dityatev et al., 2010; Kamimura and Maeda, 2017). Synaptic signals are both mediated by, and regulated via, secreted glycoproteins (GPs), proteoglycans (PGs) and endogenous glycan-binding lectins (Rushton et al., 2020). Disruption of these synaptic glycan-binding and modifying mechanisms is linked to numerous heritable neurological diseases, which include autism spectrum disorder (ASD) and intellectual disability (ID) states such as fragile X syndrome (FXS), movement and degeneration disorders such as hereditary spastic paraplegias (HSPs), and most broadly, the rapidly expanding congenital disorders of glycosylation (CDGs; Doll and Broadie, 2014; Friedman et al., 2013; Orso et al., 2005; Parkinson et al., 2016; Frappaolo et al., 2018). Secreted lectins play important roles in synaptic processes, including the secreted Ca²⁺-binding (C-type) lectins first identified by Kurt Drickamer. For example, the Drosophila Mind-the-Gap (MTG) patterns synaptomatrix glycans, regulates glutamate receptor (GluR) clustering, and modulates trans-synaptic signaling to control neurotransmission strength (Rohrbough et al., 2007; Rushton et al., 2009, 2012; Rohrbough and Broadie, 2010).

We identified a second C-type lectin, Lectin-galC1 (LGC1), in a whole-genome transgenic Drosophila RNAi screen for genes regulating synapse structure and function at the neuromuscular junction (NMJ; Dani et al., 2012). LGC1 is one of 33 C-type lectins encoded by the Drosophila genome, which are classified by their binding to extracellular β-galactoside glycans via a single characteristic Ca²⁺-dependent carbohydrate-binding domain (CBD; Ao et al., 2007; Keller and Rademacher, 2019). Similar to the many other lectins encoded by the Drosophila genome, LGC1 exists within a gene cluster with two other duplicated C-type lectin gene homologs (Theopold et al., 1999). However, these two neighbors, with homologies of 59% and 24% respectively, show both differential expression and function (Tanji et al., 2006). Previous LGC1 studies have revealed roles in development and cellular defense, showing enhanced expression in imaginal discs during pupation, binding to surface polysaccharide chains on gram-negative bacteria, and upregulation in response to injury due to an upstream nuclear factor κB-binding site, which is a common regulator of immune defense genes (Haq et al., 1996; Sun and Faye, 1992). Consistent with other glycosylation mechanisms, LGC1 has a pleiotropic role in neurotransmission (Scott and Panin, 2014). Our RNAi screen identification of LGC1 indicated critical roles in the neuromusculature, and specifically at NMJ synapses, with increased neurotransmission strength following LGC1 knockdown (Dani et al., 2012). Our goal here was to pursue this lead by generating selective tools to test LGC1 requirements in coordinated movement and NMJ neurotransmission mechanisms.

In this study, we generated a LGC1 loss-of-function null allele using CRISPR/Cas9 genome editing (Gokcezade et al., 2014) by creating a frame-shift near the beginning of the coding sequence. To characterize neuromuscular roles, we combine this null with a characterized LGC1 deficiency. We generated a LGC1::GFP fusion and specific LGC1 antibody to reveal LGC1 protein at the NMJ, in presynaptic boutons and secreted into the extracellular synaptomatrix. Consistent with this NMJ localization, we found that loss of LGC1 caused faster locomotory muscle peristalsis and much more rapid coordinated movement. Previously, we used LGC1 RNAi to show increased NMJ excitatory junction current (EJC) transmission, but no change in synaptic architecture (Dani et al., 2012). Here, we confirm these preliminary findings in null mutants employing confocal imaging and two-electrode voltage-clamp (TEVC) electrophysiology. Although the loss of LGC1 did not alter NMJ architecture, it did increase synapse number. Consistently, LGC1 mutants showed higher spontaneous synaptic vesicle (SV) fusion rates as well as elevated motor nerve stimulation-evoked EJC amplitudes. To test mechanisms, we employed FM1-43 dye imaging to reveal faster SV cycling dynamics with smaller SV pools in the mutants. With high-frequency stimulation (HFS), we found faster depression and slower recovery. Both confocal and electron microscopy revealed smaller membrane-associated SV pools at mutant active zones, consistent with faster activity-dependent SV cycling. Taken together, these results indicate that LGC1 plays a role in presynaptic SV release dynamics to cap NMJ neurotransmission strength and thus limit motor function.

In the Drosophila genome, the C-type lectin gene cluster at cytological location 37D encodes three secreted Ca²⁺-binding lectins, including Lectin-galC1 (LGC1), CG33532 and CG33533 (Dodd and Drickamer, 2001; Haq et al., 1996). A comparison of the amino acid sequences of these gene products reveals 59% similarity between LGC1 and CG33532, but a lower similarity (24%) between LGC1 and CG33533. All three genes contain two exons, with the second encoding the defining C-type lectin fold (Fig. 1A); but the products have different expression profiles and cell functions (Tanji et al., 2006). The Bloomington Drosophila Stock Center Df(2L)Exel6043 (hereafter designated ‘Df’) completely removes the LGC1 locus, C-type lectin gene cluster and other surrounding genes (Fig. 1A, red bar). To determine specific roles of LGC1 at the NMJ, we created a frame-shift (fs) complete loss-of-function null mutant allele using CRISPR/Cas9 gene editing tools (see Materials and Methods; Fig. 1A, red arrow). This mutation (designated as LGC1^fs) generates numerous substitutions and deletions, leading to a small, remnant 70-residue peptide (Fig. 1A, red text). This truncated peptide lacks any known domains, including a C-type lectin carbohydrate-binding domain (CBD). We also made a UAS-LGC1::GFP fusion transgene (hereafter designated ‘L-GFP’), inserted onto chromosome 2 at cytological location 55C4 (Fig. 1B). Both LGC1 mutants and GFP-tagged LGC1 were first used to examine neural and neuromuscular protein expression profiles.

To test LGC1 protein levels, we performed western blots of brain lysates using a new LGC1 antibody (Fig. 1B). Antibodies were generated against the entire sequence, excluding only the signal peptide (see Materials and Methods; Trier et al., 2019). The LGC1 protein has a predicted molecular mass of ∼20 kDa, based on the amino acid sequence prior to post-translational modification. Anti-LGC1 western blots show a clear protein band at ∼24 kDa in the genetic background control (w¹¹¹⁸), which is absent in both homozygous LGC1^fs/LGC1^fs and LGC1^fs/Df mutants (Fig. 1B, vertical arrows). This indicates LGC1^fs is a protein null, as predicted (Fig. 1A). Ubiquitous daughterless Gal4 (UH1-Gal4; Brand and Perrimon, 1993) driving the UAS-LGC1::GFP transgene (UH1-Gal4>L-GFP) generates a GFP fusion protein with a predicted molecular mass of ∼50 kDa. In addition to the endogenous LGC1 protein band, this line has a unique band at ∼52 kDa (Fig. 1B, horizontal arrow). Anti-GFP western blots confirm the presence of the LGC1::GFP fusion protein, and also confirm the specificity of the anti-LGC1 antibody (Fig. 1C, red arrow). There are also always several smaller GFP+ bands present in the anti-GFP western blots (horizontal arrows), which range in apparent molecular mass from ∼28–38 kDa (Fig. 1C). These bands are assumed to be GFP+ transgene degradation products. These results confirm the LGC1::GFP fusion protein and indicate that LGC1^fs is a null mutant allele.

To assess LGC1 neuromuscular expression, we first used the anti-LGC1 antibody with NMJ synaptic marker co-labeling. To test LGC1 secretion, anti-LGC1 antibody (green) was used in detergent-free labeling of larval NMJs, co-labeled with anti-horseradish peroxidase (HRP, blue) to reveal motor neuron membranes (Fig. 1D–F). LGC1 is enriched in NMJ boutons, with additional diffuse labeling surrounding boutons, owing to LGC1 protein secretion (Fig. 1D). LGC1 mutants (LGC1^fs/LGC1^fs and LGC1^fs/Df) do not show neuromuscular labeling or enrichment at the NMJ, consistent with null mutation (Fig. 1E,F). To confirm LGC1 secretion from the well-defined signal peptide, we show 24B-Gal4>L-GFP larvae secrete the LGC1::GFP fusion protein into the salivary gland (sg) lumen (Fig. 1G). Control 24B-Gal4>eGFP larvae show native GFP florescence within sg cells, but no secretion or accumulation in the sg lumen (Fig. 1H). To better localize LGC1 NMJs, animals were co-labeled with the presynaptic active zone marker anti-Bruchpilot (BRP; Fig. 1I; Marrus et al., 2004). LGC1 is enriched in presynaptic boutons (HRP, green) and overlaps with punctate active zones (BRP, magenta), with less LGC1 protein (blue) in NMJ terminal areas lacking synapses (Fig. 1I). Compared with the postsynaptic scaffold marker Discs Large (DLG, Fig. 1J; Thomas et al., 1997), LGC1 (white) is enriched in and adjacent to synaptic boutons, well within DLG-marked (blue) subsynaptic reticulum (Fig. 1J). These results show LGC1 is in the presynaptic NMJ boutons, secreted, present in the extracellular space immediately around boutons, and fully contained within the encircling postsynaptic domain.

Since NMJs mediate neurotransmission from motor neurons to muscles to drive muscle contraction, we hypothesized that LGC1 could influence coordinated movement. We previously identified LGC1 as a negative regulator of NMJ neurotransmission based on a genomic transgenic RNAi screen (Dani et al., 2012). LGC1 RNAi knockdown causes a significant increase in motoneuron stimulation-evoked NMJ transmission, but it was not at all clear how this might translate to changes in the overall animal movement function. Muscle contraction can be accurately quantified (Chandel et al., 2019). Drosophila larval body segments contain a defined motor circuit and muscle pattern, with contraction and relaxation generating peristaltic waves that propel the larva through the environment (Fushiki et al., 2016; Gjorgjieva et al., 2013). NMJ mechanisms underlie the sequential temporal and spatial activation of discrete muscles in larvae that mediate coordinated movement (Grillner, 2003). Locomotory peristaltic waves are one measurement of basal motor function driving normal movement (Nichols et al., 2012; Pulver et al., 2015). A simple roll-over assay that requires sequential muscle control to right a larva from an inverted position measures more complex, coordinated movement (Bodily et al., 2001; Jumbo-Lucioni et al., 2016). We use both approaches to compare genetic background control (w¹¹¹⁸), homozygous LGC1^fs/LGC1^fs and LGC1^fs/Df mutants. Representative still image frames from locomotion and rollover videos are shown in Fig. 2.

Movement peristalsis was measured in free-moving wandering third-instar larvae during the linear phase of locomotion, with the contraction wave proceeding sequentially along the entire body length (Fig. 2A, arrows). In w¹¹¹⁸ controls, this peristaltic wave takes just over 1 s to travel the body length (Fig. 2B). In LGC1 mutants (LGC1^fs and LGC1^fs/Df), the waves proceed in a significantly (P<0.001) more rapid segmental sequence to drive quicker forward movement, compared to slower peristaltic waves in matched controls (Fig. 2B). Coordinated movement was measured using the rollover assay on manipulated third-instar larvae, which requires a rapid sequence of bilaterally coordinated timed muscle contractions to allow the larvae to turn from an inverted position to a dorsal upwards position (Fig. 2C). In w¹¹¹⁸ controls, this righting behavior takes 9.98±1.00 s (mean±s.e.m.) on average (Fig. 2D). In contrast, LGC1^fs and LGC1^fs/Df null mutants exhibit much quicker coordinated movement that enables a very rapid righting behavior (LGC1^fs, 2.92±0.20 s; LGC1^fs/Df, 4.84±0.37 s; P<0.0001, n=30; Fig. 2D). These results indicate that LGC1 limits both peristaltic muscle waves during basal locomotion and coordinated movement during complex motor behaviors. Given LGC1 localization at the NMJ, these phenotypes suggest changes in NMJ muscle control in LGC1 mutants, owing to alterations in NMJ connectivity/architecture, synapse number and/or transmission properties from altered synaptic signaling properties.

The presence of the very significant behavioral phenotype of highly accelerated coordinated movement led us to hypothesize possible increased synapse structure or function at the NMJ. Over-elaboration in the number of NMJ synaptic boutons has been causally associated with more rapid movement (Jumbo-Lucioni et al., 2014, 2016). Likewise, increased NMJ synapse number (presynaptic active zones and postsynaptic glutamate receptors) elevates neurotransmission to drive faster muscular movement (Kittel et al., 2006; Rasse et al., 2005; Wagh et al., 2006). To test these possibilities, we assayed for changes in NMJ structure and synapse number in LGC1 mutants versus matched controls. To test structure, NMJs were co-labeled for presynaptic anti-HRP and postsynaptic anti-DLG (Fig. 3). In w¹¹¹⁸ controls, NMJs show a consistent architecture of regularly spaced synaptic boutons (Fig. 3A). In the LGC1-null mutants (LGC1^fs and LGC1^fs/Df), there is no detectable change in this usual NMJ architecture (Fig. 3B,C). In quantified measurements, there is no significant difference in NMJ area, branching or synaptic bouton number (Fig. 3D–F). These findings are consistent with our previous RNAi results, which also show no changes in NMJ area, branching or synaptic bouton number with LGC1 RNAi knockdown (Dani et al., 2012). Therefore, LGC1 has no detectable role in synaptic morphological development, but it remained possible that LGC1 may have a role in regulating synapse number within the normal NMJ boutons.

To address this possibility, we assayed LGC1-null mutants (LGC1^fs and LGC1^fs/Df) for changes in NMJ synapse number. Wandering third-instar larvae were double-labeled for presynaptic and postsynaptic components. The presynaptic active zone (AZ) marker anti-Bruchpilot (BRP) was used to mark synaptic vesicle (SV) release sites in NMJ boutons (Fig. 3G–L; Wagh et al., 2006). Postsynaptic glutamate receptor (GluR) domains were marked with an antibody against a subunit present in all these receptors (anti-GluRIIC) to label clusters in the muscle membrane (Fig. 3G–L; Marrus and DiAntonio, 2004). In w¹¹¹⁸ controls, each presynaptic AZ directly apposes a postsynaptic GluR cluster to mediate fast NMJ neurotransmission (Fig. 3G,J; Schuster et al., 1991). Synapse density, measured as synapse number per bouton area is 1.60±0.08/µm² (mean±s.e.m.; Fig. 3M). In comparison, LGC1 loss caused an increase in synapse density, with an elevated number of BRP and GluRIIC puncta (Fig. 3H,I,K,L). Both homozygous null LGC1^fs/LGC1^fs and the LGC1^fs/Df exhibit a small but significant (P<0.05) increase in synapse number per NMJ bouton area (LGC1^fs, 1.81±0.05/µm²; n=12, P=0.03; LGC1^fs/Df, 1.87±0.06/µm²; n=12, P=0.01; Fig. 3M). This increase alone is unlikely to explain the much more significant elevation in coordinated movement in LGC1 mutants. Moreover, although structural and functional differentiation can be related, they are independent processes (Menon et al., 2013). Therefore, we next assayed whether LGC1 plays a role in NMJ neurotransmission function.

Our RNAi screen indicated that LGC1 knockdown elevates neurotransmission strength (Dani et al., 2012). However, RNAi knockdown only causes a partial loss of function, and there is always the concern of possible off-target RNAi effects (Seinen et al., 2011; Heigwer et al., 2018). We therefore next tested the contributions of LGC1 to NMJ function using electrophysiological recordings in our new CRISPR/Cas9-generated null mutant alone and in trans to deficiency. A two-electrode voltage-clamp (TEVC) configuration was used to make both motor nerve stimulation-evoked excitatory junction current (EJC) and spontaneous miniature EJC (mEJC) recordings using 1.0 mM [Ca²⁺] physiological saline (Parkinson et al., 2016; Kopke et al., 2020). For mEJC recordings, the motor nerve was severed and the muscle was voltage-clamped at −60 mV, with continuous records made in 2-min sessions. For evoked EJCs, a suction electrode was used to stimulate the motor nerve innervating the muscle at a suprathreshold voltage (Rohrbough et al., 1999). Repeated recordings (0.2 Hz, 10 consecutive stimuli) were averaged to determine the mean peak EJC amplitude for each animal (N=1). TEVC records were made from staged wandering third-instar larvae comparing the genetic background control (w¹¹¹⁸) to the homozygous LGC1^fs/LGC1^fs null mutants and LGC1^fs/Df. Representative evoked EJC and spontaneous mEJC traces, as well as their quantified measurements, are all shown in Fig. 4.

Nerve stimulation results in robust NMJ glutamatergic transmission. In the w¹¹¹⁸ controls, muscle TEVC records show a reliable, high fidelity EJC following each motor nerve stimulation (Fig. 4A). The control EJC amplitude averaged from 10 consecutive recordings is 154.8±7.4 nA (mean±s.e.m.; Fig. 4B). Both LGC1-null mutants (LGC1^fs/LGC1^fs and LGC1^fs/Df) exhibited clearly larger EJC responses (Fig. 4A, right), consistent with our earlier LGC1 RNAi results (Dani et al., 2012). Quantification reveals similar evoked EJC transmission amplitudes of 184.8±8.8 nA for LGC1^fs and 176.3±7.2 nA for LGC1^fs/Df (Fig. 4B). The amplitudes were not significantly different from each other, but both LGC1 nulls were significantly increased compared with genetic background controls (Fig. 4B). The mutant phenotype reflects a selective elevation in the peak EJC neurotransmission amplitude without any detectable changes in EJC kinetics, including stimulation latency, EJC rise time and decay constant (Fig. 4A). An increased EJC indicates a strengthened NMJ due to either greater quantal content, with more SVs released during evoked nerve stimulation (Pan and Zucker, 2009), or a greater number/function of the postsynaptic glutamate receptors (Astorga et al., 2016; Gerkin et al., 2013), or both together. To distinguish these possibilities, we next assayed spontaneous mEJCs, to test presynaptic vesicle fusion probability (mEJC frequency), as well as the number of activated postsynaptic receptors (mEJC amplitude).

We used mEJC recordings to begin to dissect the presynaptic and postsynaptic mechanisms regulated by LGC1 (Bykhovskaia and Vasin, 2017). Representative mEJC traces from all three genotypes are shown in Fig. 4C. In the w¹¹¹⁸ controls, mEJC events occur at a regular low frequency and amplitude (Fig. 4C, top). Quantification reveals an average control mEJC frequency of 0.50±0.07 Hz, with an average control mEJC amplitude of 0.32±0.06 nA (mean±s.e.m.; Fig. 4D,E). In contrast, both LGC1-null mutants (LGC1^fs/LGC1^fs and LGC1^fs/Df) exhibit an obvious increase in mEJC frequency, albeit without any apparent change in the mEJC amplitude (Fig. 4C, middle and bottom). Quantification reveals that mEJC frequency is significantly higher (P<0.001) in both LGC1^fs (1.94±0.30 Hz) and LGC1^fs/Df (2.07±0.29 Hz), without any change in either of the mutant mEJC amplitudes (Fig. 4D,E). These results indicate that LGC1 limits evoked neurotransmission and spontaneous vesicle release. Therefore, the LGC1-null effects are primarily presynaptic, with an increase in SV fusion probability that elevates evoked quantal content in the absence of LGC1 function. The increased evoked EJC amplitude in LGC1-null mutants perhaps reflects the greater synapse number (Fig. 3J–M), but also likely increased SV release probability at each synapse (Fig. 4A–D). Moreover, the functional strengthening suggests additional effects on SV dynamics, which we therefore next analyzed using live optical imaging.

The above findings indicate that LGC1 function limits neurotransmission strength, due to an elevated presynaptic fusion probability. Since SV cycling dynamics drive presynaptic vesicle availability and therefore their release capabilities (Gan and Watanabe, 2018; Chanaday et al., 2019), we next tested LGC1 roles in the SV cycle underlying functional neurotransmission. We employed the activity-dependent, depolarization-driven lipophilic FM1-43 dye labeling technique to monitor the SV cycle (Betz and Bewick, 1992), which is widely used to quantitatively measure SV endocytosis and exocytosis at the Drosophila NMJ (Kopke and Broadie, 2018). In this approach, the wandering third-instar larva neuromusculature is bathed in FM1-43 dye (4 µM), which is incorporated into cycling SVs (‘loaded’) through endocytosis after stimulation with a motor nerve suction electrode as above (Kopke et al., 2017). This loading provides a quantitative measurement of the functional SV cycling pool size (Staples and Broadie, 2013; Vijayakrishnan et al., 2009). A second electrical stimulation of the motor nerve causes depolarization in the absence of FM1-43 dye, driving SV fusion and release of the FM1-43 dye (‘unloaded’) through exocytosis. This provides a measurement of SV release efficacy from presynaptic active zones. Representative low- and high-magnification images comparing FM1-43 dye loading and unloading in genetic background control (w¹¹¹⁸), and the LGC1^fs and LGC1^fs/Df null mutants are shown in Fig. 5.

In w¹¹¹⁸ control NMJs, boutons load FM1-43 dye upon motor nerve stimulation, with boutons loaded strongly throughout the synaptic terminal (Fig. 5A) and intense fluorescence within individual presynaptic boutons (Fig. 5G). In LGC1-null mutants, there is significantly less dye incorporation throughout the entire NMJ (Fig. 5C,E) and within individual synaptic boutons (Fig. 5I,K). Compared to the control FM1-43 uptake (56.6±4.9, mean±s.e.m.), both LGC1^fs (32.1±6.6) and LGC1^fs/Df (24.0±3.1) exhibit very significantly less dye loading (Fig. 5M). The w¹¹¹⁸ control NMJs unload most of the FM1-43 upon the second stimulation (Fig. 5B), although some of the dye is still retained within individual NMJ boutons (Fig. 5H). In contrast, the LGC1-null mutants show negligible amounts of dye remaining after unloading (Fig. 5D,F). Importantly, both LGC1-null boutons retain much less FM1-43 than control boutons following unloading (Fig. 5J,L). The quantified measurements show LGC1 nulls display significantly (P<0.01) elevated SV exocytosis, releasing more dye from their presynaptic boutons (w¹¹¹⁸ control, 22.8±2.8; LGC1^fs, 9.2±2.3; LGC1^fs/Df, 7.5±1.8; Fig. 5M). Thus, both LGC1^fs and LGC1^fs/Df mutants appear to have smaller SV cycling pools, with much less of the FM1-43 dye taken up, but a faster rate of SV cycling, with higher levels of exocytosis. These results suggest a likely increase in synaptic depression during high frequency usage owing to a smaller SV pool and faster activity-dependent SV cycling.

The mismatch in SV pool size versus release rate suggests LGC1-null NMJs will fatigue more rapidly (Fergestad et al., 1999; Rudling et al., 2018). At the Drosophila NMJ, high-frequency neurotransmission challenges the synaptic release machinery, especially at high-probability release sites, which exhibit activity-dependent depression (Kauwe and Isacoff, 2013). Following high activity levels, new SV generation and SV recruitment enables recovery of neurotransmission amplitudes within tens of seconds (Bui and Glavinović, 2014). We therefore hypothesized that the LGC1 mutant SV cycling defects determined with FM1-43 imaging (Fig. 5) would cause faster activity-dependent synaptic depression as well as slower recovery from this depression. To test this idea, variable high frequency stimulation (HFS; 5, 20 and 50 Hz; Stevens et al., 2012; Kauwe and Isacoff, 2013; Newman et al., 2017) in conjunction with TEVC recording was used to measure both synaptic depression and recovery (Fig. 6A). The HFS paradigm used low frequency (0.2 Hz) stimulation to establish the EJC amplitude baseline, following by HFS for 5 min (5, 20 or 50 Hz) to measure EJC amplitude depression. In addition to studying depression, the following EJC amplitude recovery was measured at 30 s intervals at basal 0.2 Hz stimulation (Fig. 6A). Sample TEVC traces of the genetic background control (w¹¹¹⁸) compared to LGC1^fs and LGC1^fs/Df mutants, together with detailed frequency quantifications, are all shown in Fig. 6.

Representative EJC recordings from all three genotypes at specified HFS time points (t=0, 1 s, 2 min and 5 min) reveal the relative depression of synaptic transmission over time (Fig. 6B). After 30 s of low frequency rest, the fifth trace (+30 s) shows the relative partial recovery of EJC amplitude in the different genotypes (Fig. 6B). The control and LGC1 nulls all show progressive presynaptic depression with increasing HFS frequencies (5–50 Hz; Fig. 6C). However, LGC1^fs and LGC1^fs/Df show significantly decreased EJC amplitudes immediately following the 5 Hz HFS train initiation (Fig. 6C). LGC1 loss causes greater SV release in <1 s with HFS, consistent with the greater SV release probability in nulls (Fig. 6D). At higher frequencies (20–50 Hz), significantly greater synaptic depression no longer occurs in LGC1 nulls, presumably due to heightened activity-dependent demand depressing controls beyond tolerance (Fig. 6C). Controls and LGC1 nulls all show progressive reduced recovery with increasing HFS frequencies (5–50 Hz; Fig. 6C, right, red shading). However, both LGC1^fs and LGC1^fs/Df have significantly slower recovery rates. Following all HFS trains (5–50 Hz), LGC1 nulls show reduced recovery at 30 s normalized to the starting basal EJC amplitudes (5 Hz, w¹¹¹⁸, 0.88±0.03; LGC1^fs, 0.53±0.08; 20 Hz, w¹¹¹⁸, 0.69±0.05; LGC1^fs, 0.41±0.04; and 50 Hz, w¹¹¹⁸, 0.59±0.07; LGC1^fs, 0.27±0.07; Fig. 6D). These functional defects suggest LGC1 loss alters presynaptic SV availability, which we next attempted to image.

In order to visualize presynaptic SV distribution in control boutons compared to LGC1-null mutants, we combined confocal imaging for SV pools (Daniels et al., 2004; Martin and Krantz, 2014) with imaging via transmission electron microscopy (TEM) ultrastructure studies to directly image individual SVs (Dear et al., 2016; Kopke and Broadie, 2018). We first used immunocytochemistry to label NMJ synaptic boutons with anti-vesicular glutamate transporter (VGLUT), an integral vesicle membrane protein that specifically labels all glutamatergic SVs (Wilson et al., 2005). Confocal imaging of this marker provides a quantitative measurement of SV number within entire NMJ boutons, as well as some means to map SV distribution relative to the presynaptic membrane in controls compared to LGC1^fs and LGC1^fs/Df (Fig. 7). TEM studies reveal the synaptic bouton ultrastructure, including the number and distribution of presynaptic active zones, SV density relative to active zones, and the muscle membrane subsynaptic reticulum (SSR) architecture (Wagner, 2017). We quantified all of these synaptic ultrastructural features in genetic background control (w¹¹¹⁸) compared to LGC1^fs null mutants (Fig. 8). However, the focus of both the light and electron microscopy imaging was on SV density and SV distribution within individual NMJ synaptic boutons, to provide better mechanistic insights into the movement behavior and functional NMJ neurotransmission changes characterizing the LGC1-null condition.

In w¹¹¹⁸ control NMJs, VGLUT-marked SVs are restricted to synaptic boutons in a dense cortical ring immediately underlying the plasma membrane (Fig. 7A, arrow inset). In LGC1 nulls, this labeling is clearly and consistently reduced (Fig. 7B,C). Compared to the VGLUT intensity in controls (49.6±2.5, mean±s.e.m.), the mean intensity is significantly (P<0.05) decreased in both the LGC1^fs (40.6±2.2) and LGC1^fs/Df (41.8±2.4) mutants (Fig. 7D). This signal loss suggests a lower SV density within NMJ boutons, which we confirmed with TEM imaging (Fig. 8). To examine the SV distribution within individual boutons, higher magnification VGLUT distribution measurements were compared between genotypes. LGC1 nulls show a relatively uniform, lower intensity fluorescence within NMJ boutons, while controls have increased VGLUT near the presynaptic membrane (Fig. 7E–G). In w¹¹¹⁸ controls, line graphs of fluorescence intensity show a clear decrease in signal in the bouton interior with peaks associated with each membrane (Fig. 7E,H). In sharp contrast, both the LGC1^fs and LGC1^fs/Df null mutants show a relatively uniform VGLUT signal distribution across the entire synaptic bouton, and lack the two membrane peaks (Fig. 7F,G,I,J). Note that the lowest fluorescence values for all three genotypes are similar, suggesting that mutants are selectively deficient in SVs near the membrane. These results indicate altered SV distributions in mutants at the whole-bouton level, but this would be better supported by SV imaging at single vesicle resolution.

TEM studies show almost no synaptic ultrastructural changes in LGC1 nulls, with normal synaptic boutons, SVs, AZs and postsynaptic SSR compared to w¹¹¹⁸ controls (Fig. 8A–D). To examine SV distribution, numbers were quantified within three consecutive 100-nm diameter semicircular zones at AZs (Fig. 8B,D, yellow zones 1–3). Studies show a decrease in SV density across all zones in the LGC1 mutants (Fig. 8D). Quantification shows both controls and the LGC1^fs null exhibit a decrease in SV density from zone 1 to 3, with mutants showing lower density in all zones. The SV density decrease in zone 1 and 2 is particularly significant in LGC1^fs nulls (P<0.0001; Fig. 8E). In individual TEM slices, SV density within all three zones is significantly reduced in LGC1^fs mutants versus controls, with the difference increasing as the AZ membrane is approached (Fig. 8F). In the closest zone 1, the control density is 665±56 SV/µm² (mean±s.e.m.), compared to 468±28 SV/µm² in the LGC1-null mutants (Fig. 8F). Taken together with the above results, we suggest LGC1 loss elevates synaptic function by leading to greater SV release at an increased number of synapses. Increased neurotransmission translates to accelerated movement, causing rapid peristaltic waves and faster coordinated motion. However, strengthened NMJs also exhibit an impairment in replenishing activity-depleted SV pools in mutants. Thus, LGC1 regulates movement and NMJ synaptic function via activity-dependent changes in both SV cycling dynamics and SV release probability.

Anterograde and retrograde signals secreted into the NMJ synaptomatrix by both motor neuron and muscle cells orchestrate intercellular communication at the synapse (Rushton et al., 2020). The signaling ligands, their receptors and other molecules in the synaptomatrix environment are all very highly glycosylated, and mutation of many genes affecting glycosylation has strong effects on both coordinated movement and NMJ synaptic function (Broadie et al., 2011; Dani et al., 2012, 2014; Jumbo-Lucioni et al., 2014, 2016; Parkinson et al., 2013, 2016). Glycan-binding lectins interact with the NMJ synaptomatrix at many levels, to effectively modulate NMJ neurotransmission strength. In the extracellular space, lectins serve to regulate the distribution of glycans (as scaffolds, co-receptors and signaling platforms), thereby directing secreted and transmembrane protein function (Del Fresno et al., 2018; Sych et al., 2018; Choi et al., 2019). For example, we have extensively characterized the secreted C-type lectin Mind-the-Gap (MTG) at the Drosophila NMJ, which regulates glycan distribution, secreted trans-synaptic signaling ligands, synaptic position-specific (PS) integrins and the postsynaptic glutamate receptor (GluR) field, as well as the downstream internal Drosophila Pix–Pak–Dock signaling cascade and the critical DLG synaptic scaffold (Rohrbough et al., 2007; Rushton et al., 2009, 2012; Rohrbough and Broadie, 2010). These secreted MTG C-type lectin roles controlling postsynaptic function suggest that similar C-type lectin roles may govern presynaptic function.

In a previous systematic transgenic RNAi screen for NMJ modulator genes, we did indeed uncover a new secreted C-type lectin that strongly controls neurotransmission strength (Dani et al., 2012). Lectin-galC1 (LGC1) is an ideal candidate for modulation of synaptic function due to its characterized Ca²⁺-binding (C-type) and glycan-binding properties as a secreted protein (Tanji et al., 2006; Haq et al., 1996). To create a LGC1-specific null mutation, we here used CRISPR/Cas9 homology-directed DNA repair to create a targeted frame-shift mutation (Fig. 1A–C; Gratz et al., 2013). We paired this LGC1^fs null mutant with a deletion removing the C-type lectin gene cluster containing LGC1 to characterize a new specific LGC1 antibody and to show that LGC1 is locally secreted at the NMJ (Fig. 1D–J). We then moved forward to study the effects of LGC1 loss on coordinated movement, which intimately depends on both NMJ synaptic architecture and function. Consistent with our initial discovery of increased NMJ neurotransmission following LGC1 RNAi knockdown (Dani et al., 2012), and earlier studies demonstrating LGC1 Ca²⁺-dependent binding functions (Tanji et al., 2006; Haq et al., 1996), we report here that LGC1-null mutants exhibit accelerated movement (Fig. 2). LGC1 loss increases both the rate of locomotion and the speed to complete complex coordinated behaviors when challenged. These findings suggest LGC1 acting locally at the NMJ is important for modulating locomotion and, particularly, complex coordinated motor function.

LGC1 is secreted at the NMJ, as demonstrated using non-permeabilized labeling (Rushton et al., 2012; Kopke et al., 2017), and concentrated within the extracellular synaptomatrix surrounding synaptic boutons; a space characterized by striking glycan accumulation (Parkinson et al., 2013; Rushton et al., 2020). LGC1-null mutants have absolutely no alterations in NMJ structure (Fig. 3A–F). This is surprising, since NMJ overelaboration is a common feature of glycosylation mutations (Rushton et al., 2020). For example, mutations in the key phosphomannomutase type 2 (PMM2) glycosylation enzyme strongly alter synaptomatrix pauci-mannose glycans to cause significant NMJ architectural expansion, including more synaptic branching and supernumerary boutons (Parkinson et al., 2016). Similarly, loss of UDP-sugar precursors, which are required for synaptic glycosylation also results in obvious NMJ structural overelaboration, with both increased branching and synaptic bouton number (Jumbo-Lucioni et al., 2014, 2016). However, synaptic structure and function are independently regulated, with separable glycan roles (Collins and DiAntonio, 2007; Miller et al., 2012). Consistently, LGC1-null mutants exhibit an elevated number of synapses within otherwise normal NMJ boutons, thereby exhibiting greater synaptic density (Fig. 3G–M). LGC1 nulls have more presynaptic BRP-positive active zones (AZs) and more apposed postsynaptic GluR fields, suggesting that a higher level of synaptic connectivity contributes to elevated neurotransmission strength.

LGC1 limits NMJ function by reducing SV fusion and evoked transmission (Fig. 4). An elevated LGC1 EJC amplitude reflects an increase in number of released vesicles (quantal content; Perry et al., 2020) in a more mobile readily releasable pool (RRP; Vaden et al., 2019). Since spontaneous mEJC amplitude does not change in LGC1 nulls (Fig. 4), an increase in postsynaptic GluR number or function is ruled out as a contributing factor (Bykhovskaia and Vasin, 2017; Kaeser and Regehr, 2017). In addition to increased synapse density, we hypothesized that there would be higher SV fusion probability from altered SV cycling in LGC1 mutants. FM1-43 dye imaging (Rodesch and Broadie, 2000; Kopke and Broadie, 2018) shows that LGC1 loss results in a smaller cycling SV pool with higher cycling rate (Fig. 5). A smaller SV pool could negate the effects of increased synapse number and SV fusion probability. However, most SVs do not participate detectably in transmission cycling, and smaller SV pools can show high transmission due to increased synaptic density and SV cycling rates (Denker et al., 2011). Moreover, altered SV endocytosis–exocytosis cycling efficacy can cause altered synaptic depression/recovery with high-frequency stimulation (HFS; Stevens et al., 2012). Stimulation frequency can change SV cycling dynamics, with higher frequencies triggering differential SV release and recovery mechanisms (Maeno-Hikichi et al., 2011). Consequently, we assayed NMJ function (depression and recovery) during varying levels of frequency stimulation, to stress neurotransmission machinery, and thus gain insight into the LGC1 mutant SV cycling mechanisms.

In LGC1 nulls, HFS causes more rapid fatigue and slower recovery (Fig. 6). Consistent with this, smaller SV cycling pools with higher turnover rates can increase basal transmission strength while impairing transmission maintenance during HFS demand (Baldelli et al., 2007; Doussau et al., 2017). Thus, the smaller cycling SV pool in LGC1 mutants (Fig. 5) can maintain elevated basal transmission while exhibiting impaired replenishment of SV pools and maintenance of transmission during HFS (Fig. 6). Other glutamatergic synapse classes maintain a small cycling RRP for high probability SV release during basal stimulation conditions, yet are unable to maintain transmission under conditions of greater demand (Xu-Friedman et al., 2001; Valera et al., 2012). Likewise, LGC1 nulls exhibit reduced membrane-associated SV pools at both the light microscope (Fig. 7) and electron microscope (Fig. 8) levels, including lower SV density at AZs. Other Drosophila mutants with increased NMJ neurotransmission can similarly show presynaptic SV depletion in TEM ultrastructural studies (Jia et al., 1993). At the AZ, SVs near the presynaptic membrane make up the RRP and linked cycling pools (Moulder and Mennerick, 2005), but internal SV pools can be held in ‘reserve’ or serve other functions apart from neurotransmitter release (Denker et al., 2011). Overall, our findings are consistent with LGC1 mutants exhibiting a smaller and faster SV cycling pool, with altered SV cycle dynamics during and following HFS stress.

Similar to what is seen with LGC1 mutants, the loss of a number of C-type lectin domain (CTLD) proteins causes synaptic defects. For example, Mincle-deficient mice suppress TNF induction, which lowers glutamate release from central synapses, resulting in a direct modulation of neurotransmission strength (Richardson and Williams, 2014; Ishikawa et al., 2019). Although LGC1 appears to have the opposite effect, the Drosophila TNF homolog (Eiger) downregulates excitatory amino acid transporters 1 and 2 (EAAT1/2; Igaki et al., 2009), and the EAAT1 mutation impairs locomotory peristaltic movement (Stacey et al., 2010). We show here that LGC1 limits locomotory peristaltic movement, suggesting a possible relationship between LGC1, Eiger and EAAT1 may be a useful avenue to explore in future studies. Another CTLD protein is Caenorhabditis elegans CLEC-38, which has been shown to regulate presynaptic organization and mediate changes in the integral SV protein Synaptobrevin (Kulkarni et al., 2008). As the vesicle SNARE (V-SNARE), Synaptobrevin drives SV fusion with the presynaptic membrane AZ to directly mediate neurotransmitter release (Haberman et al., 2012; Estes et al., 2000). Since LGC1 mutants display neurotransmission defects comparable to CLEC-38 mutants, such as similar increased SV fusion and impaired SV cycling, the exploration of downstream LGC1 interactions and Synaptobrevin regulation in the presynaptic terminal could offer more insight into the mechanisms by which LGC1 produces neurotransmission changes.

In conclusion, this study provides insight into the roles of a novel C-type lectin at a glutamatergic neuromuscular synapse. Beyond the previously characterized roles in development and immunity (Tanji et al., 2006; Haq et al., 1996), we find LGC1 secreted at the NMJ synapse modulates presynaptic function. Along with acting to limit synapse number and neurotransmission strength, LGC1 also regulates SV cycling to increase SV availability during basal levels of use, but decrease SV availability during high levels of use. In resting basal conditions, LGC1 caps NMJ neurotransmission strength to limit coordinated motor function. With intense stimulation, LGC1 loss causes more fatigue and decreases the recovery rate, indicating impaired activity-dependent SV cycling. We reveal functional defects with electrophysiology, and visualize SVs with both FM1-43 dye and TEM electron microscopy imaging. In future studies, we plan to dissect LGC1 mechanisms by focusing on SV pools and trafficking, as well as testing the possible involvement in signaling via Eiger, EAAT1 and downstream SV Synaptobrevin changes. Our work indicates that LGC1-dependent signaling fine-tunes the multiple presynaptic pools that exchange SVs differentially based on activity levels and release demand to control SV release probability (Bui and Glavinović, 2014; Guo et al., 2015; Millar et al., 2002). Beyond this LGC1 mechanism, we hope this new model will allow further exploration of SV cycling-dependent NMJ function in neuromuscular disease states.

All stocks were reared on standard cornmeal, molasses, yeast and agar food medium under standard conditions of constant 25°C with a 12 h light–12 h dark cycle. The genetic background control was w¹¹¹⁸. The deficiency Df(2 L)Exel6043 (BSC#7525) removes the entire lectin-galC1 (LGC1) gene, as well as two downstream homologs and several upstream genes. A CRISPR/Cas9-generated LGC1 frame-shift mutant (LGC1^fs) was made and confirmed (described below). Mutants were assayed as LGC1^fs/LGC1^fs and LGC1^fs/Df(2L)Exel6043. A UAS-LGC1::GFP line was made with LGC1 cDNA attached to GFP coding sequence with a flexible linker. This construct was injected into y¹w^67c23; P{CaryP}attP1 (BDSC#8621) embryos to insert on chromosome 2 at cytological location 55C4 (Bloomington Drosophila Stock Center). The ubiquitous Gal4 driver used was for the daughterless (da) gene (P{Gal4-da.G32}UH1; UH1-Gal4; Wodarz et al., 1995).

Both mutagenesis and mutant line isolation were conducted as previously described (Shilts and Broadie, 2017). Briefly, a frame-shift (fs) null LGC1 allele (LGC1^fs) was generated using CRISPR/Cas9 genome editing (Gratz et al., 2013). The guide RNA (gRNA) oligonucleotide sequence targeting the LGC1 gene was:

5′-CACTGAAATGCTGAAGCTTAGTTT-3′.

The gRNA was assembled in a pCFD3-dU6:3gRNA vector (Addgene #49410; Port et al., 2014). Purified constructs were injected (250 ng/μl) into w¹¹¹⁸, vas-Cas9 Drosophila embryos (BestGene, Inc.). Injected embryos were then crossed into the homogenized genetic background. PCR sequencing was used on isolated mutant alleles to verify nucleotide sequence changes. One characterized frame-shift null allele (LGC1^fs) was used in all the phenotypic studies. For phenotype comparisons, LGC1^fs was crossed with the molecularly defined Df(2L)Exel6043 deficiency to generate LGC1^fs/Df.

A new LGC1 antibody was generated against amino acids 21–186 of LGC1 (ABClonal, Woburn, Massachusetts, USA). This amino acid sequence represents the entire LGC1 protein, excluding only the signal peptide. The specific sequence used was: REKFSIQVNEGNTFGALVKAEPFTKINDGYYFFGTESLNWYEAYEKCRELNSELVTFETDQEFDAVTAFLTANGSRLTYWTSGNDLAKTGSHRWFTNAQRISSLRWARNQPDNAGQKEHCIHLGYIYKDSRKFELNDRPCSQDPNSLFKYICEAPEMETISIVVWK.

The sequence was determined to be antigenic and display no cross-reactivity with the Drosophila LGC1 homologs (CG33532 and CG33533). Three antiserums (16, 17, 18) were recovered and affinity-purified from three separate injected rabbits. LGC1 Ab16 was preabsorbed against the LGC1^fs null mutants prior to use in this study.

Western blotting was performed as previously described (Vita and Broadie, 2017). Briefly, two adult males at 1 day post-eclosion (dpe) were homogenized in 1× NuPAGE LDS sample buffer (45 μl, Invitrogen, NP007), heated at 95°C with 2-mercaptoethanol (2-ME; 5 μl, Sigma-Aldrich, M7154) for 10 min, and then centrifuged at 7800 g for 10 min. Supernatant (10 μl) was loaded onto 4–12% Bis-Tris gels (Invitrogen, NP0336) and run with NuPage 1× MES running buffer (Invitrogen, NP002). Samples were transferred to nitrocellulose membranes (PROTRAN, NBA085C001EA) in NuPage transfer buffer with 20% methanol at 30 mA. Membranes were labeled for protein content with Ponceau S stain (Sigma-Aldrich, P7170). Odyssey blocking buffer diluted in TBS-T (10 mM Tris-HCl pH 8, 150 mM NaCl and 0.05% Tween 20) was used to block membranes for 1 h at room temperature (RT). Samples were incubated in rabbit anti-LGC1 (1:5000) and goat anti-GFP (1:2500; ab6662; Abcam) for 2 h at RT. Membranes were subsequently washed with TBS-T and incubated in Alexa Fluor 680 donkey anti-rabbit-IgG (1:10,000; Jackson Laboratory) and Alexa Fluor 800 donkey anti-goat-IgG (1:10,000; Jackson Laboratory). Blots were imaged with an Odyssey CLx infrared scanner (Licor).

Immunolabeling of wandering third-instar larvae was performed as previously described (Staples and Broadie, 2013). Briefly, staged animals were dissected in physiological saline (128 mM NaCl, 2 mM KCl, 4 mM MgCl₂, 0.2 mM CaCl₂, 70 mM sucrose, 5.5 mM trehalose, 5 mM HEPES, pH 7.2). A lateral cut was made at the posterior and extended anteriorly in a dorsal longitudinal incision, and the body walls were pinned down. The internal organs were removed, revealing body wall muscles, central nervous system (CNS) and motor nerves. The preparations were fixed in 4% paraformaldehyde (PFA) for 10 min, and then blocked with 0.5% bovine serum albumin (BSA) for 10 min. Preparations were either permeabilized with 0.2% Triton X-100 (for HRP, DLG, BRP, GluRIIC and VGLUT), or processed without detergent for extracellular only labeling (LGC1). Primary antibodies used were: rabbit anti-LGC1 (1:50; this study), mouse anti-Discs Large [DLG; 1:250; Developmental Studies Hybridoma Bank (DSHB), 4F3], goat anti-horse radish peroxidase (HRP) Alexa Fluor647-conjugated (1:300; Jackson Laboratories, 123-605-021), mouse anti-Bruchpilot (BRP; 1:100; DSHB, nc82), rabbit anti-glutamate receptor IIC (GluRIIC; 1:5000, Marrus and DiAntonio, 2004), and rabbit anti-vesicular glutamate receptor (VGLUT; 1:10,000; Daniels et al., 2004). Preparations were incubated in primary antibody overnight at 4°C. Secondary antibodies used were: goat anti-rabbit-IgG Alexa Fluor 488-conjugated, goat anti-mouse-IgG Alexa Fluor 555-conjugated, goat anti-mouse-IgG Alexa Fluor488-conjugated, and donkey anti-rabbit-IgG Alexa Fluor 555-conjugated, all at 1:300 (Invitrogen). Preparations were incubated in secondary antibodies for 2 h at RT. Preparations were mounted in Fluoromount G (Electron Microscopy Services) for confocal imaging.

Laser-scanning confocal microscopy was used for all imaging as previously described (Sears et al., 2019). Briefly, 40× water-immersion (for FM1-43) and 63× oil-immersion (for all immunocytochemistry) Plan Apo objective lens were used for imaging on a Zeiss LSM 510 META microscope. Muscle 4 NMJs in abdominal segments 3 and 4 were imaged in Z-stacks. All fluorescence quantifications were undertaken using ImageJ software (NIH, open source). Maximum intensity projections were made using the ‘Z Project Function’. Regions of interest (ROIs) were determined at the NMJ with the anti-HRP:633 channel using the ‘Threshold Function’. ROIs were subsequently applied to other channels, with the ‘Measure Function’ used to obtain average fluorescence intensities. The gray value intensity line graphs were created using the ‘Plot Profile Function’. BRP puncta were quantified using the ‘Find Maxima Function’. For NMJ structural analyses, type Ib synaptic boutons were defined as HRP- and DLG-positive varicosities with a diameter ≥2 μm. Branches were defined as containing ≥2 boutons.

Peristaltic contraction was quantified as reported previously (Shilts and Broadie, 2017). Briefly, an individual wandering third-instar larva was placed at the center of a 2% agarose plate and acclimated for 30 s at RT. With even illumination, peristaltic waves were video recorded using an Olympus dissection microscope and mounted Canon Rebel DLSR camera. Uninterrupted movement (1 min) was recorded for each animal. Videos were analyzed at 0.5× speed on Windows MPC-HC. Peristaltic motion was defined as the complete movement of a muscular contraction wave from posterior to anterior. Five complete waves were measured and averaged per animal for a single data point (N=1). Coordinated movement roll-over assays were performed as previously described (Jumbo-Lucioni et al., 2016). Briefly, an individual wandering third-instar larva was placed on a 2% agarose plate and acclimated for 30 s at RT. A fine paintbrush was used to gently roll the larva until the ventral midline was exactly upwards (t=0). A stopwatch was then used to record the time until the ventral middle was exactly downwards. The assay was repeated three times on the same animal, and the three times were then averaged for each single data point (N=1).

The two-electrode voltage-clamp (TEVC) electrophysiology configuration was employed as previously described (Kopke et al., 2020). Briefly, wandering third-instar larvae were dissected along the dorsal midline and the walls glued down (3M Vetbond Tissue Adhesive). The internal organs were removed to reveal the neuromusculature. The peripheral nerves were cut at the CNS exit points. Recordings were carried out at 18°C in physiological saline (see above). A 40× water-immersion objective lens on a Zeiss Axioskop microscope was used to image the preparation. Two microelectrodes (1-mm outer diameter borosilicate capillaries; World Precision Instruments) were inserted into muscle 6 of abdominal segments 3 or 4. Microelectrodes were ∼15 MΩ resistance when filled with 3 M KCl. The muscle was voltage-clamped at −60 mV (Axoclamp-2B amplifier). Spontaneous miniature excitatory junction current (mEJC) recordings were made in continuous 2 min sessions and then low-pass filtered. For nerve stimulation-evoked EJC recordings, a fire-polished glass suction electrode was used to suck up the motor nerve. A Grass (S88) stimulator was used to stimulate the motor nerve with 0.5 ms suprathreshold voltage stimuli at 0.2 Hz. EJC recordings were filtered at 2 kHz. To quantify EJC amplitude, 10 consecutive traces were averaged and the average peak value recorded. For high-frequency stimulation (HFS) train assays, motor nerves were stimulated at 5 Hz, 20 Hz or 50 Hz for 5 min. For recovery assays, ten consecutive EJC traces with 0.2 Hz stimulation were averaged at 30 s intervals following the HFS train. Clampex 9.0 was used for data acquisition, and Clampfit 9 was used for data analyses.

FM1-43 dye labeling was done as previously described (Kopke and Broadie, 2018). Briefly, staged wandering third-instar larvae were dissected in physiological saline (see above) and NMJ terminals labeled with anti-HRP:647 (100 μl; Sigma) added to the bath. Peripheral nerves were cut at the CNS, and the CNS removed. The bath solution was then replaced with FM1-43 (4 μM; Invitrogen) in 1.0 mM [Ca²⁺] physiological saline. As above, a glass suction electrode was used to suck up the motor nerve for stimulation (20 Hz, 5 min). The bath solution was then replaced several times in quick succession with Ca²⁺-free saline to halt synaptic vesicle (SV) cycling. Z-stacks of the FM1-43-loaded NMJ were taken with the Zeiss LSM 510 confocal microscope using a 40× water immersion objective. Next, the Ca²⁺-free saline was replaced with 1.0 mM [Ca²⁺] saline (without FM1-43) and the same motor nerve stimulated (20 Hz, 1 min) to drive SV exocytosis and dye release. The bath solution was then replaced several times in quick succession with Ca²⁺-free saline to halt SV cycling. Z-stacks of the same NMJ were then taken. Images were quantified with ImageJ (NIH) by setting a threshold using the anti-HRP:633 channel to highlight the NMJ. The FM1-43 dye channel was then selected, and the ‘Measure Function’ used to assay fluorescence intensity.

Synaptic transmission electron microscopy (TEM) was undertaken as previously described (Kopke et al., 2017). Briefly, staged wandering third-instar larvae were dissected as above and fixed overnight at 4°C in 2.5% glutaraldehyde. A secondary fixation was then performed in 1% osmium tetroxide for 1 h at RT. The preparations were washed in 0.1 M sodium cacodylate buffer three times (10 min each), and then ddH₂O three times (15 min each). En bloc 2% uranyl acetate staining was undertaken for 2 h at RT in the dark. Preparations were again rinsed in ddH₂O three times (15 min each), followed by an ethanol dehydration series (30, 50, 70, 90, 95, 100, 100%), propylene oxide infiltration, and resin embedding (Embed-812). Muscles 6 or 7 from abdominal segments 2 or 3 were dissected free and embedded in a semi-hardened resin block. Isolated muscles from four animals were put in each block, with three blocks for each genotype. Blocks were polymerized at 60°C for 48 h. Thick sections (1 µm) were cut from the blocks using a glass knife, stained with Toluidine Blue (Sigma-Aldrich) for 1 min on a Thermostat slide warmer (45°C), and then imaged on a compound microscope with a 100× objective (total magnification 1000×) for NMJ bouton identification. Once a bouton was found, ultrathin (50 nm) sections were cut using a DiATOME diamond knife on a Leica Ultracut UCT ultramicrotome, and then collected on uncoated 200 mesh copper grids. All TEM imaging was performed on a Philips CM10 transmission electron microscope operating at 80 kV, with images collected using a 4-megapixel AMT CCD camera. Bouton area, SSR area and SV size, number and distribution were all measured using ImageJ.

All study statistics were performed using Prism software (GraphPad, San Diego, CA). All data sets were subjected to Shapiro–Wilks normality tests. All comparisons between any two experimental groups were performed by using the two-tailed Welch's unequal variances t-test, or the Student's two-tailed unpaired t-test, as indicated. All comparisons between ≥3 experimental groups were performed by using one-way ANOVA tests, except when the paired data were assayed, for which two-way ANOVA tests were used, as indicated. All display graphs were made using Prism software to show all the individual data points (N indicated) as well as the mean±s.e.m. In all of the figures, significance is indicated as *P≤0.05, **P≤0.01, ***P≤0.001, ****P<0.0001 and P>0.05 (not significant, ns).

We are grateful to Daniel Okoye and Darius Booth for their technical input on this study. We thank both the Bloomington Drosophila Stock Center (BDSC; Indiana University, Bloomington, IN, USA) and the Vienna Drosophila Resource Center (VDRC; Vienna, Austria) for providing essential genetic lines for this study. We thank the Developmental Studies Hybridoma Bank (DSHB; University of Iowa, Iowa City, IA, USA) for providing essential antibodies. We are especially grateful to Dr Aaron DiAntonio (Washington University, St Louis, MO, USA) for the gift of both GluRIIC and VGLUT antibodies.